Because of the lack of any meaningful information on deletions, the only way for people who don't work for StackExchange to do useful statistical analysis on question closures (to answer questions like "Just how badly is the close queue backlog affecting close speed?") is
live-scraping the real site using the StackExchange API. Otherwise, it's too biased. You end up with crazy results - like that the average time for closing a 5 year old Java question is about 5 years, and question closing gets monotonically faster and faster every month!
So are there any types of closed questions that are unlikely to be deleted?
What about closes as duplicates? Duplicates are useful, right? At least, we used to think so.